Real-time Full-plane Magnetic Field Deep Integration Observation System Based On GPU Technology

High-resolution solar magnetic field observation is a goal that solar physicists unremittingly pursue.The use of modern image processing techniques and methods to process late data can make full use of the potential of existing telescopes.For ground-based telescopes,the resolution of the telescope is determined by the coherent scale of the Earth's atmospheric disturbance.The solar magnetic field requires long time integral observations,and it is also affected by wind,telescope tracking error and other influencing factors,resulting in much lower resolution than the ideal coherence scale.The observation of full-disk solar magnetic field generally uses multiple full-disk solar images to deeply integrate to reduce the magnetic noise and improve the signal-to-noise ratio,so as to obtain a higher-quality full-disk solar magnetograms.Therefore,due to the deep integration of the image,the computer needs to align each full-disk solar image for cross-correlation.Performing image alignment and cross-correlational operations necessarily involves the delineation of cross-correlational windows,but since the image gradient features of full-disk solar images have large differences in the presence or absence of sunspots,we use different cross-correlational window demarcation strategies to perform image cross-correlational alignment of full-disk solar images.In addition,in order to meet the real-time requirements of the observers for the software,this paper uses GPU technology to design and implement a full-disk solar magnetic field real-time observation system.This system mainly realizes the real-time displacement integral calculation for the images acquired by the camera.The specific work is mainly divided into the following two points:First,image processing techniques are used to denoise the initial data obtained,and gradient maps are used to determine the existence and location of sunspots.Different cross-correlational strategies are implemented: global cross-correlation or local cross-correlation.and the use of cubic spline interpolation for cross-correlational results.The sub-pixel solution is used to improve the experimental accuracy.Finally,the magnetic field data is calculated according to the calculation formula of the full-disk solar magnetograms.Second,using the CUDA(Compute Unified Device Architecture)framework to use GPU parallel computing to improve and implement the above algorithm,greatly improving the efficiency of the algorithm to meet the current real-time requirements.This paper takes the Huairou Solar Observatory of the National Astronomical Observatory of the Chinese Academy of Sciences as its research and application background.The main achievements are as follows: First,a cross-correlation alignment algorithm for full-disk solar magnetic field data is designed and implemented.The results of image quality evaluation demonstrate the effectiveness of the algorithm.Second,the above-mentioned cross-correlation alignment algorithm is implemented using GPU parallel acceleration technology,the real-time requirements of system was realized.The current work results are being tested at Huairou Solar Observatory,the National Astronomical Observatory of the Chinese Academy of Sciences.